hallucination

A hallucination is when an AI model produces information that is false or entirely made up, but presents it with the same confidence as a correct answer. Because a large language model generates plausible-sounding text rather than looking up facts, it can invent citations, statistics, or even web addresses that never existed.

This is not a rare glitch but a built-in characteristic of how these models work, and it has real-world consequences. In one example, attackers noticed that AI tools tend to invent the same fake web addresses repeatedly, so they registered those addresses and set up malicious sites, waiting for users to arrive.

The practical defence is simple: verify anything that matters — facts, figures, links and quotes — before relying on it. Treat an AI’s output as a confident draft, not a finished source of truth.